Re: [PATCH v8 2/2] s390/kvm: diagnose 0x318 sync and reset

[Collin Walling <walling@linux.ibm.com>]

On 6/19/20 10:52 AM, David Hildenbrand wrote:
> On 19.06.20 00:22, Collin Walling wrote:
>> DIAGNOSE 0x318 (diag318) sets information regarding the environment
>> the VM is running in (Linux, z/VM, etc) and is observed via
>> firmware/service events.
>>
>> This is a privileged s390x instruction that must be intercepted by
>> SIE. Userspace handles the instruction as well as migration. Data
>> is communicated via VCPU register synchronization.
>>
>> The Control Program Name Code (CPNC) is stored in the SIE block. The
>> CPNC along with the Control Program Version Code (CPVC) are stored
>> in the kvm_vcpu_arch struct.
>>
>> The CPNC is shadowed/unshadowed in VSIE.
>>
> 
> [...]
> 
>>  
>>  int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
>> @@ -4194,6 +4198,10 @@ static void sync_regs_fmt2(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
>>  		if (vcpu->arch.pfault_token == KVM_S390_PFAULT_TOKEN_INVALID)
>>  			kvm_clear_async_pf_completion_queue(vcpu);
>>  	}
>> +	if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
>> +		vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
>> +		vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
>> +	}
>>  	/*
>>  	 * If userspace sets the riccb (e.g. after migration) to a valid state,
>>  	 * we should enable RI here instead of doing the lazy enablement.
>> @@ -4295,6 +4303,7 @@ static void store_regs_fmt2(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
>>  	kvm_run->s.regs.pp = vcpu->arch.sie_block->pp;
>>  	kvm_run->s.regs.gbea = vcpu->arch.sie_block->gbea;
>>  	kvm_run->s.regs.bpbc = (vcpu->arch.sie_block->fpf & FPF_BPBC) == FPF_BPBC;
>> +	kvm_run->s.regs.diag318 = vcpu->arch.diag318_info.val;
>>  	if (MACHINE_HAS_GS) {
>>  		__ctl_set_bit(2, 4);
>>  		if (vcpu->arch.gs_enabled)
>> diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
>> index 9e9056cebfcf..ba83d0568bc7 100644
>> --- a/arch/s390/kvm/vsie.c
>> +++ b/arch/s390/kvm/vsie.c
>> @@ -423,6 +423,8 @@ static void unshadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>>  		break;
>>  	}
>>  
>> +	scb_o->cpnc = scb_s->cpnc;
> 
> "This is a privileged s390x instruction that must be intercepted", how
> can the cpnc change, then, while in SIE?
> 
> Apart from that LGTM.
> 

I thought shadow/unshadow was a load/store (respectively) when executing
in SIE for a level 3+ guest (where LPAR is level 1)?

* Shadow SCB (load shadow VSIE page; originally CPNC is 0)

* Execute diag318 (under SIE)

* Unshadow SCB (store in original VSIE page; CPNC is whatever code the
guest decided to set)

Don't we need to preserve the CPNC for the level 3+ guest somehow?

-- 
Regards,
Collin

Stay safe and stay healthy